next up previous
Next: Tunnelling through two dots Up: Incoherent tunnelling through two Previous: The canonical distribution

Tunnelling through a single dot

 

When the energy level broadening is negligible compared to the temperature, i.e. in the limit of weak coupling to the reservoirs, the density of states in the dot can be adequately described by a set of delta functions. The leads are assumed to be in thermal equilibrium, described by the Fermi-Dirac distributions tex2html_wrap_inline1264 and tex2html_wrap_inline1266. To a first approximation, the electron-electron interaction can satisfactorily be described using the charging model, where each pair of electrons has an associated Coulomb repulsion of U. Fermi's golden rule gives the tunnelling rates between the dot and the reservoirs for all energy levels k.
eqnarray247
where tex2html_wrap_inline1272 is the strength of coupling to the left reservoir. The expressions for the tunnelling rates through the right barrier are similar. It is assumed that the quantum dot can be approximated by a parabolic confining potential so that the single particle energy levels are equally spaced by an energy tex2html_wrap_inline1274. In figures 2 and 3 the current and its associated differential conductance is plotted as a function of the applied bias for a range of energy level spacings. In numerical calculations it is possible to take into account more realistic energy level spectra, enabling a closer comparison with experiment. When it is taken into account that the total spin of the system can only change by 1/2 with each tunnelling event, it appears that negative differential conductance may occur in specific regions [14, 15].

In general a dot with closely packed energy levels yields a higher current than a dot with a sparse energy level spectrum, because of the higher number of current paths available. When tex2html_wrap_inline1278 then the metallic regime is entered.

When the energy level spacing is not negligible, the I-V characteristics are typified by two energy scales, the Coulomb repulsion energy U and the bare energy level spacing tex2html_wrap_inline1274. As in the metallic regime, one expects the I-V characteristic to display a current step whenever the maximum occupation of the dot increases by one. This happens with a period tex2html_wrap_inline1284, since an extra electron not only has to overcome the Coulomb barrier but also has to tunnel to the next available energy level. In addition, there is also some fine structure which has an associated period of tex2html_wrap_inline1274. This is caused by the fact that an extra current path is created when the bias is increased by tex2html_wrap_inline1274.

When tex2html_wrap_inline1290 the dot will be mostly maximally occupied and the most marked current increases occur whenever the Coulomb blockade can be overcome. This means that the period tex2html_wrap_inline1284 is accentuated. In the opposite regime tex2html_wrap_inline1294 the dominant period is the level spacing. Under normal operating conditions tex2html_wrap_inline1296 the two periods coexist (see figure 3). It is noted that at higher bias voltages the number of peaks in the differential conductance increases, as new current paths become available at different energies for different occupation numbers. Only when the ratio tex2html_wrap_inline1298 is an integer do peaks corresponding to different occupation numbers coincide.

The Ohmic conductance through a quantum dot (figure 4) differs from that through a metallic dot in two significant ways. Firstly, the periodicity of the conductance peaks has increased by the level spacing tex2html_wrap_inline1274. Secondly, the temperature dependence of the peaks has changed its nature. An increase in temperature now not only leads to larger thermal broadening, but also to a lowering of the peak amplitude which is inversely proportional to the temperature. This is due to the fact that transport proceeds through a single energy level. The temperature dependence is proportional to the derivative of the Fermi-distribution function in the reservoirs [16]. Therefore, at low temperatures tex2html_wrap_inline1302 the Ohmic conductance can be written as
equation268
where the successive charge degeneracy points tex2html_wrap_inline1304 are spaced by an energy tex2html_wrap_inline1284. For higher temperatures several levels should be taken into consideration, each of which is weighted by a factor determined by the Boltzmann distribution [10].
 equation275
where tex2html_wrap_inline1308 is the joint probability that the dot contains N electrons and that the single particle level tex2html_wrap_inline1312 is empty. Note that the contributions from levels other than at positions tex2html_wrap_inline1304 are too small to cause a peak in the conductance.

Apart from the I-V characteristics and the Ohmic conductance, there is a third useful experiment that can be carried out. This involves applying a constant bias across the dot and studying the resultant current as a function of the gate potential. Typically the source-drain bias is chosen to be less than the charging energy which isolates the effects of the energy separation of the zero dimensional states of the dot. In figure 5 the current has been calculated as a function of the gate voltage for a dot with a constant level spacing tex2html_wrap_inline1316. The thermal energy is given by tex2html_wrap_inline1318. In accordance with some recent experiments [17, 18, 19] a number of peaks and troughs can be observed in the current. All peaks and troughs can be characterised by the number of levels available to an incoming electron to tunnel onto, and the number of levels from which an electron can tunnel out of the dot. For a Fermi level separation tex2html_wrap_inline1320 of 0.15 U between the reservoirs (see figure 5) these numbers are tex2html_wrap_inline1324. This explains why the total peak is split into two subpeaks. For the second set of graphs with tex2html_wrap_inline1326 the sequence of available tunnelling levels is tex2html_wrap_inline1328. It is also clear that asymmetric tunnelling barriers cause the current peaks to be asymmetric.


next up previous
Next: Tunnelling through two dots Up: Incoherent tunnelling through two Previous: The canonical distribution

Angus MacKinnon - Aonghus Mac Fhionghuin Fri Nov 8 18:10:54 gmt 1996